Large file support + memory usage review

[dkarlovi]

https://phpspreadsheet.readthedocs.io/en/develop/topics/reading-files/#reading-only-specific-columns-and-rows-from-a-file-read-filters

[dkarlovi]

See PHPOffice/PhpSpreadsheet#629 (comment)

[ohaag]

Hello,
I don't know if this is the good place since I don't know at all this repository.
I followed the topic PHPOffice/PhpSpreadsheet#629 which led me here and I wanted to provide some perfomance comparisons.

This was tested with the exact same code automatically migrated with rector from phpoffice/phpexcel 1.8.2 to phpoffice/phpspreadsheet 1.25.2, with XDebug disabled on a symfony prod environment.
A CSV file with 129820 lines and 20 columns (~20Mo) was processed (reader->load + same business logic every time)

with a chunk size of 3000 :

• 1.2min with phpexcel
• 5.8min with phpspreadsheet

with a chunk size of 10000 :

• 34sec with phpexcel

with a chunk size of 65000 :

• 26sec with phpexcel
• 4.0min with phpspreadsheet

Of course, I checked in debug where duration explode to more than 15min (I stopped here), the vast majority of time is spent on the calls of CsvReader->load().

A large part of the duration (4min become 1.2min) is caused by the following lines. Is it really mandatory ? If it is some aesthetic behavior could we just make it optional ?